1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
|
;******************************************************************************
;* VP8 MMXEXT optimizations
;* Copyright (c) 2010 Ronald S. Bultje <rsbultje@gmail.com>
;* Copyright (c) 2010 Fiona Glaser <fiona@x264.com>
;*
;* This file is part of FFmpeg.
;*
;* FFmpeg is free software; you can redistribute it and/or
;* modify it under the terms of the GNU Lesser General Public
;* License as published by the Free Software Foundation; either
;* version 2.1 of the License, or (at your option) any later version.
;*
;* FFmpeg is distributed in the hope that it will be useful,
;* but WITHOUT ANY WARRANTY; without even the implied warranty of
;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
;* Lesser General Public License for more details.
;*
;* You should have received a copy of the GNU Lesser General Public
;* License along with FFmpeg; if not, write to the Free Software
;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
;******************************************************************************
%include "libavutil/x86/x86util.asm"
SECTION_RODATA
pw_27: times 8 dw 27
pw_63: times 8 dw 63
pb_4: times 16 db 4
pb_F8: times 16 db 0xF8
pb_FE: times 16 db 0xFE
pb_27_63: times 8 db 27, 63
pb_18_63: times 8 db 18, 63
pb_9_63: times 8 db 9, 63
cextern pb_1
cextern pb_3
cextern pw_9
cextern pw_18
cextern pb_80
SECTION .text
;-----------------------------------------------------------------------------
; void ff_vp8_h/v_loop_filter_simple_<opt>(uint8_t *dst, ptrdiff_t stride, int flim);
;-----------------------------------------------------------------------------
; macro called with 7 mm register indexes as argument, and 5 regular registers
; first 11 mean the same as READ_8x4_TRANSPOSED above
; fifth regular register is scratchspace to reach the bottom 8 rows, it
; will be set to second regular register + 8*stride at the end
%macro READ_16x4_INTERLEAVED 12
; transpose 16 (A-P) rows of 4 pixels each
lea %12, [r0+8*r2]
; read (and interleave) those addressable by %8 (=r0), A/C/D/E/I/K/L/M
movd m%1, [%8+%10*4] ; A0-3
movd m%3, [%12+%10*4] ; I0-3
movd m%2, [%8+%10*2] ; C0-3
movd m%4, [%12+%10*2] ; K0-3
movd m%6, [%8+%10] ; D0-3
movd m%5, [%12+%10] ; L0-3
movd m%7, [%12] ; M0-3
add %12, %11
punpcklbw m%1, m%3 ; A/I
movd m%3, [%8] ; E0-3
punpcklbw m%2, m%4 ; C/K
punpcklbw m%6, m%5 ; D/L
punpcklbw m%3, m%7 ; E/M
punpcklbw m%2, m%6 ; C/D/K/L interleaved
; read (and interleave) those addressable by %9 (=r4), B/F/G/H/J/N/O/P
movd m%5, [%9+%10*4] ; B0-3
movd m%4, [%12+%10*4] ; J0-3
movd m%7, [%9] ; F0-3
movd m%6, [%12] ; N0-3
punpcklbw m%5, m%4 ; B/J
punpcklbw m%7, m%6 ; F/N
punpcklbw m%1, m%5 ; A/B/I/J interleaved
punpcklbw m%3, m%7 ; E/F/M/N interleaved
movd m%4, [%9+%11] ; G0-3
movd m%6, [%12+%11] ; O0-3
movd m%5, [%9+%11*2] ; H0-3
movd m%7, [%12+%11*2] ; P0-3
punpcklbw m%4, m%6 ; G/O
punpcklbw m%5, m%7 ; H/P
punpcklbw m%4, m%5 ; G/H/O/P interleaved
%endmacro
; write 4 xmm registers of 4 dwords each
; arguments same as WRITE_2x4D, but with an extra register, so that the 5 regular
; registers contain buf+4*stride, buf+5*stride, buf+12*stride, -stride and +stride
; we add 1*stride to the third regular registry in the process
; the 10th argument is 16 if it's a Y filter (i.e. all regular registers cover the
; same memory region), or 8 if they cover two separate buffers (third one points to
; a different memory region than the first two), allowing for more optimal code for
; the 16-width case
%macro WRITE_4x4D 10
; write out (4 dwords per register), start with dwords zero
movd [%5+%8*4], m%1
movd [%5], m%2
movd [%7+%8*4], m%3
movd [%7], m%4
; store dwords 1
psrldq m%1, 4
psrldq m%2, 4
psrldq m%3, 4
psrldq m%4, 4
movd [%6+%8*4], m%1
movd [%6], m%2
%if %10 == 16
movd [%6+%9*4], m%3
%endif
movd [%7+%9], m%4
; write dwords 2
psrldq m%1, 4
psrldq m%2, 4
%if %10 == 8
movd [%5+%8*2], m%1
movd %5d, m%3
%endif
psrldq m%3, 4
psrldq m%4, 4
%if %10 == 16
movd [%5+%8*2], m%1
%endif
movd [%6+%9], m%2
movd [%7+%8*2], m%3
movd [%7+%9*2], m%4
add %7, %9
; store dwords 3
psrldq m%1, 4
psrldq m%2, 4
psrldq m%3, 4
psrldq m%4, 4
%if %10 == 8
mov [%7+%8*4], %5d
movd [%6+%8*2], m%1
%else
movd [%5+%8], m%1
%endif
movd [%6+%9*2], m%2
movd [%7+%8*2], m%3
movd [%7+%9*2], m%4
%endmacro
%macro WRITE_8W 5
%if cpuflag(sse4)
pextrw [%3+%4*4], %1, 0
pextrw [%2+%4*4], %1, 1
pextrw [%3+%4*2], %1, 2
pextrw [%3+%4 ], %1, 3
pextrw [%3 ], %1, 4
pextrw [%2 ], %1, 5
pextrw [%2+%5 ], %1, 6
pextrw [%2+%5*2], %1, 7
%else
movd %2d, %1
psrldq %1, 4
mov [%3+%4*4], %2w
shr %2, 16
add %3, %5
mov [%3+%4*4], %2w
movd %2d, %1
psrldq %1, 4
add %3, %4
mov [%3+%4*2], %2w
shr %2, 16
mov [%3+%4 ], %2w
movd %2d, %1
psrldq %1, 4
mov [%3 ], %2w
shr %2, 16
mov [%3+%5 ], %2w
movd %2d, %1
add %3, %5
mov [%3+%5 ], %2w
shr %2, 16
mov [%3+%5*2], %2w
%endif
%endmacro
%macro SIMPLE_LOOPFILTER 2
cglobal vp8_%1_loop_filter_simple, 3, %2, 8, dst, stride, flim, cntr
%if cpuflag(ssse3)
pxor m0, m0
%endif
SPLATB_REG m7, flim, m0 ; splat "flim" into register
; set up indexes to address 4 rows
DEFINE_ARGS dst1, mstride, stride, dst3, dst2
mov strideq, mstrideq
neg mstrideq
%ifidn %1, h
lea dst1q, [dst1q+4*strideq-2]
%endif
%ifidn %1, v
; read 4 half/full rows of pixels
mova m0, [dst1q+mstrideq*2] ; p1
mova m1, [dst1q+mstrideq] ; p0
mova m2, [dst1q] ; q0
mova m3, [dst1q+ strideq] ; q1
%else ; h
lea dst2q, [dst1q+ strideq]
READ_16x4_INTERLEAVED 0, 1, 2, 3, 4, 5, 6, dst1q, dst2q, mstrideq, strideq, dst3q
TRANSPOSE4x4W 0, 1, 2, 3, 4
%endif
; simple_limit
mova m5, m2 ; m5=backup of q0
mova m6, m1 ; m6=backup of p0
psubusb m1, m2 ; p0-q0
psubusb m2, m6 ; q0-p0
por m1, m2 ; FFABS(p0-q0)
paddusb m1, m1 ; m1=FFABS(p0-q0)*2
mova m4, m3
mova m2, m0
psubusb m3, m0 ; q1-p1
psubusb m0, m4 ; p1-q1
por m3, m0 ; FFABS(p1-q1)
mova m0, [pb_80]
pxor m2, m0
pxor m4, m0
psubsb m2, m4 ; m2=p1-q1 (signed) backup for below
pand m3, [pb_FE]
psrlq m3, 1 ; m3=FFABS(p1-q1)/2, this can be used signed
paddusb m3, m1
psubusb m3, m7
pxor m1, m1
pcmpeqb m3, m1 ; abs(p0-q0)*2+abs(p1-q1)/2<=flim mask(0xff/0x0)
; filter_common (use m2/p1-q1, m4=q0, m6=p0, m5/q0-p0 and m3/mask)
mova m4, m5
pxor m5, m0
pxor m0, m6
psubsb m5, m0 ; q0-p0 (signed)
paddsb m2, m5
paddsb m2, m5
paddsb m2, m5 ; a=(p1-q1) + 3*(q0-p0)
pand m2, m3 ; apply filter mask (m3)
mova m3, [pb_F8]
mova m1, m2
paddsb m2, [pb_4] ; f1<<3=a+4
paddsb m1, [pb_3] ; f2<<3=a+3
pand m2, m3
pand m1, m3 ; cache f2<<3
pxor m0, m0
pxor m3, m3
pcmpgtb m0, m2 ; which values are <0?
psubb m3, m2 ; -f1<<3
psrlq m2, 3 ; +f1
psrlq m3, 3 ; -f1
pand m3, m0
pandn m0, m2
psubusb m4, m0
paddusb m4, m3 ; q0-f1
pxor m0, m0
pxor m3, m3
pcmpgtb m0, m1 ; which values are <0?
psubb m3, m1 ; -f2<<3
psrlq m1, 3 ; +f2
psrlq m3, 3 ; -f2
pand m3, m0
pandn m0, m1
paddusb m6, m0
psubusb m6, m3 ; p0+f2
; store
%ifidn %1, v
mova [dst1q], m4
mova [dst1q+mstrideq], m6
%else ; h
inc dst1q
SBUTTERFLY bw, 6, 4, 0
%if cpuflag(sse4)
inc dst2q
%endif
WRITE_8W m6, dst2q, dst1q, mstrideq, strideq
lea dst2q, [dst3q+mstrideq+1]
%if cpuflag(sse4)
inc dst3q
%endif
WRITE_8W m4, dst3q, dst2q, mstrideq, strideq
%endif
RET
%endmacro
INIT_XMM sse2
SIMPLE_LOOPFILTER v, 3
SIMPLE_LOOPFILTER h, 5
INIT_XMM ssse3
SIMPLE_LOOPFILTER v, 3
SIMPLE_LOOPFILTER h, 5
INIT_XMM sse4
SIMPLE_LOOPFILTER h, 5
;-----------------------------------------------------------------------------
; void ff_vp8_h/v_loop_filter<size>_inner_<opt>(uint8_t *dst, [uint8_t *v,] ptrdiff_t stride,
; int flimE, int flimI, int hev_thr);
;-----------------------------------------------------------------------------
%macro INNER_LOOPFILTER 2
%define stack_size 0
%ifndef m8 ; stack layout: [0]=E, [1]=I, [2]=hev_thr
%ifidn %1, v ; [3]=hev() result
%define stack_size mmsize * -4
%else ; h ; extra storage space for transposes
%define stack_size mmsize * -5
%endif
%endif
%if %2 == 8 ; chroma
cglobal vp8_%1_loop_filter8uv_inner, 6, 6, 13, stack_size, dst, dst8, stride, flimE, flimI, hevthr
%else ; luma
cglobal vp8_%1_loop_filter16y_inner, 5, 5, 13, stack_size, dst, stride, flimE, flimI, hevthr
%endif
%if cpuflag(ssse3)
pxor m7, m7
%endif
%ifndef m8
; splat function arguments
SPLATB_REG m0, flimEq, m7 ; E
SPLATB_REG m1, flimIq, m7 ; I
SPLATB_REG m2, hevthrq, m7 ; hev_thresh
%define m_flimE [rsp]
%define m_flimI [rsp+mmsize]
%define m_hevthr [rsp+mmsize*2]
%define m_maskres [rsp+mmsize*3]
%define m_p0backup [rsp+mmsize*3]
%define m_q0backup [rsp+mmsize*4]
mova m_flimE, m0
mova m_flimI, m1
mova m_hevthr, m2
%else
%define m_flimE m9
%define m_flimI m10
%define m_hevthr m11
%define m_maskres m12
%define m_p0backup m12
%define m_q0backup m8
; splat function arguments
SPLATB_REG m_flimE, flimEq, m7 ; E
SPLATB_REG m_flimI, flimIq, m7 ; I
SPLATB_REG m_hevthr, hevthrq, m7 ; hev_thresh
%endif
%if %2 == 8 ; chroma
DEFINE_ARGS dst1, dst8, mstride, stride, dst2
%else
DEFINE_ARGS dst1, mstride, stride, dst2, dst8
%endif
mov strideq, mstrideq
neg mstrideq
%ifidn %1, h
lea dst1q, [dst1q+strideq*4-4]
%if %2 == 8 ; chroma
lea dst8q, [dst8q+strideq*4-4]
%endif
%endif
; read
lea dst2q, [dst1q+strideq]
%ifidn %1, v
%if %2 == 8 && mmsize == 16
%define movrow movh
%else
%define movrow mova
%endif
movrow m0, [dst1q+mstrideq*4] ; p3
movrow m1, [dst2q+mstrideq*4] ; p2
movrow m2, [dst1q+mstrideq*2] ; p1
movrow m5, [dst2q] ; q1
movrow m6, [dst2q+ strideq*1] ; q2
movrow m7, [dst2q+ strideq*2] ; q3
%if mmsize == 16 && %2 == 8
movhps m0, [dst8q+mstrideq*4]
movhps m2, [dst8q+mstrideq*2]
add dst8q, strideq
movhps m1, [dst8q+mstrideq*4]
movhps m5, [dst8q]
movhps m6, [dst8q+ strideq ]
movhps m7, [dst8q+ strideq*2]
add dst8q, mstrideq
%endif
%else ; h
%if %2 == 16
lea dst8q, [dst1q+ strideq*8]
%endif
; read 16 rows of 8px each, interleave
movh m0, [dst1q+mstrideq*4]
movh m1, [dst8q+mstrideq*4]
movh m2, [dst1q+mstrideq*2]
movh m5, [dst8q+mstrideq*2]
movh m3, [dst1q+mstrideq ]
movh m6, [dst8q+mstrideq ]
movh m4, [dst1q]
movh m7, [dst8q]
punpcklbw m0, m1 ; A/I
punpcklbw m2, m5 ; C/K
punpcklbw m3, m6 ; D/L
punpcklbw m4, m7 ; E/M
add dst8q, strideq
movh m1, [dst2q+mstrideq*4]
movh m6, [dst8q+mstrideq*4]
movh m5, [dst2q]
movh m7, [dst8q]
punpcklbw m1, m6 ; B/J
punpcklbw m5, m7 ; F/N
movh m6, [dst2q+ strideq ]
movh m7, [dst8q+ strideq ]
punpcklbw m6, m7 ; G/O
; 8x16 transpose
TRANSPOSE4x4B 0, 1, 2, 3, 7
%ifdef m8
SWAP 1, 8
%else
mova m_q0backup, m1
%endif
movh m7, [dst2q+ strideq*2]
movh m1, [dst8q+ strideq*2]
punpcklbw m7, m1 ; H/P
TRANSPOSE4x4B 4, 5, 6, 7, 1
SBUTTERFLY dq, 0, 4, 1 ; p3/p2
SBUTTERFLY dq, 2, 6, 1 ; q0/q1
SBUTTERFLY dq, 3, 7, 1 ; q2/q3
%ifdef m8
SWAP 1, 8
SWAP 2, 8
%else
mova m1, m_q0backup
mova m_q0backup, m2 ; store q0
%endif
SBUTTERFLY dq, 1, 5, 2 ; p1/p0
%ifdef m12
SWAP 5, 12
%else
mova m_p0backup, m5 ; store p0
%endif
SWAP 1, 4
SWAP 2, 4
SWAP 6, 3
SWAP 5, 3
%endif
; normal_limit for p3-p2, p2-p1, q3-q2 and q2-q1
mova m4, m1
SWAP 4, 1
psubusb m4, m0 ; p2-p3
psubusb m0, m1 ; p3-p2
por m0, m4 ; abs(p3-p2)
mova m4, m2
SWAP 4, 2
psubusb m4, m1 ; p1-p2
psubusb m1, m2 ; p2-p1
por m1, m4 ; abs(p2-p1)
mova m4, m6
SWAP 4, 6
psubusb m4, m7 ; q2-q3
psubusb m7, m6 ; q3-q2
por m7, m4 ; abs(q3-q2)
mova m4, m5
SWAP 4, 5
psubusb m4, m6 ; q1-q2
psubusb m6, m5 ; q2-q1
por m6, m4 ; abs(q2-q1)
pmaxub m0, m1
pmaxub m6, m7
pmaxub m0, m6
; normal_limit and high_edge_variance for p1-p0, q1-q0
SWAP 7, 3 ; now m7 is zero
%ifidn %1, v
movrow m3, [dst1q+mstrideq ] ; p0
%if mmsize == 16 && %2 == 8
movhps m3, [dst8q+mstrideq ]
%endif
%elifdef m12
SWAP 3, 12
%else
mova m3, m_p0backup
%endif
mova m1, m2
SWAP 1, 2
mova m6, m3
SWAP 3, 6
psubusb m1, m3 ; p1-p0
psubusb m6, m2 ; p0-p1
por m1, m6 ; abs(p1-p0)
pmaxub m0, m1 ; max_I
SWAP 1, 4 ; max_hev_thresh
SWAP 6, 4 ; now m6 is I
%ifidn %1, v
movrow m4, [dst1q] ; q0
%if mmsize == 16 && %2 == 8
movhps m4, [dst8q]
%endif
%elifdef m8
SWAP 4, 8
%else
mova m4, m_q0backup
%endif
mova m1, m4
SWAP 1, 4
mova m7, m5
SWAP 7, 5
psubusb m1, m5 ; q0-q1
psubusb m7, m4 ; q1-q0
por m1, m7 ; abs(q1-q0)
pxor m7, m7
pmaxub m0, m1
pmaxub m6, m1
psubusb m0, m_flimI
psubusb m6, m_hevthr
pcmpeqb m0, m7 ; max(abs(..)) <= I
pcmpeqb m6, m7 ; !(max(abs..) > thresh)
%ifdef m12
SWAP 6, 12
%else
mova m_maskres, m6 ; !(abs(p1-p0) > hev_t || abs(q1-q0) > hev_t)
%endif
; simple_limit
mova m1, m3
SWAP 1, 3
mova m6, m4 ; keep copies of p0/q0 around for later use
SWAP 6, 4
psubusb m1, m4 ; p0-q0
psubusb m6, m3 ; q0-p0
por m1, m6 ; abs(q0-p0)
paddusb m1, m1 ; m1=2*abs(q0-p0)
mova m7, m2
SWAP 7, 2
mova m6, m5
SWAP 6, 5
psubusb m7, m5 ; p1-q1
psubusb m6, m2 ; q1-p1
por m7, m6 ; abs(q1-p1)
pxor m6, m6
pand m7, [pb_FE]
psrlq m7, 1 ; abs(q1-p1)/2
paddusb m7, m1 ; abs(q0-p0)*2+abs(q1-p1)/2
psubusb m7, m_flimE
pcmpeqb m7, m6 ; abs(q0-p0)*2+abs(q1-p1)/2 <= E
pand m0, m7 ; normal_limit result
; filter_common; at this point, m2-m5=p1-q1 and m0 is filter_mask
%ifdef m8 ; x86-64 && sse2
mova m8, [pb_80]
%define m_pb_80 m8
%else ; x86-32 or mmx/mmxext
%define m_pb_80 [pb_80]
%endif
mova m1, m4
mova m7, m3
pxor m1, m_pb_80
pxor m7, m_pb_80
psubsb m1, m7 ; (signed) q0-p0
mova m6, m2
mova m7, m5
pxor m6, m_pb_80
pxor m7, m_pb_80
psubsb m6, m7 ; (signed) p1-q1
mova m7, m_maskres
pandn m7, m6
paddsb m7, m1
paddsb m7, m1
paddsb m7, m1 ; 3*(q0-p0)+is4tap?(p1-q1)
pand m7, m0
mova m1, [pb_F8]
mova m6, m7
paddsb m7, [pb_3]
paddsb m6, [pb_4]
pand m7, m1
pand m6, m1
pxor m1, m1
pxor m0, m0
pcmpgtb m1, m7
psubb m0, m7
psrlq m7, 3 ; +f2
psrlq m0, 3 ; -f2
pand m0, m1
pandn m1, m7
psubusb m3, m0
paddusb m3, m1 ; p0+f2
pxor m1, m1
pxor m0, m0
pcmpgtb m0, m6
psubb m1, m6
psrlq m6, 3 ; +f1
psrlq m1, 3 ; -f1
pand m1, m0
pandn m0, m6
psubusb m4, m0
paddusb m4, m1 ; q0-f1
%ifdef m12
SWAP 6, 12
%else
mova m6, m_maskres
%endif
pxor m7, m7
pand m0, m6
pand m1, m6
psubusb m1, [pb_1]
pavgb m0, m7 ; a
pavgb m1, m7 ; -a
psubusb m5, m0
psubusb m2, m1
paddusb m5, m1 ; q1-a
paddusb m2, m0 ; p1+a
; store
%ifidn %1, v
movrow [dst1q+mstrideq*2], m2
movrow [dst1q+mstrideq ], m3
movrow [dst1q], m4
movrow [dst1q+ strideq ], m5
%if mmsize == 16 && %2 == 8
movhps [dst8q+mstrideq*2], m2
movhps [dst8q+mstrideq ], m3
movhps [dst8q], m4
movhps [dst8q+ strideq ], m5
%endif
%else ; h
add dst1q, 2
add dst2q, 2
; 4x8/16 transpose
TRANSPOSE4x4B 2, 3, 4, 5, 6
lea dst8q, [dst8q+mstrideq +2]
WRITE_4x4D 2, 3, 4, 5, dst1q, dst2q, dst8q, mstrideq, strideq, %2
%endif
RET
%endmacro
INIT_XMM sse2
INNER_LOOPFILTER v, 16
INNER_LOOPFILTER h, 16
INNER_LOOPFILTER v, 8
INNER_LOOPFILTER h, 8
INIT_XMM ssse3
INNER_LOOPFILTER v, 16
INNER_LOOPFILTER h, 16
INNER_LOOPFILTER v, 8
INNER_LOOPFILTER h, 8
;-----------------------------------------------------------------------------
; void ff_vp8_h/v_loop_filter<size>_mbedge_<opt>(uint8_t *dst, [uint8_t *v,] ptrdiff_t stride,
; int flimE, int flimI, int hev_thr);
;-----------------------------------------------------------------------------
%macro MBEDGE_LOOPFILTER 2
%define stack_size 0
%ifndef m8 ; stack layout: [0]=E, [1]=I, [2]=hev_thr
%if mmsize == 16 ; [3]=hev() result
; [4]=filter tmp result
; [5]/[6] = p2/q2 backup
; [7]=lim_res sign result
%define stack_size mmsize * -7
%else ; 8 ; extra storage space for transposes
%define stack_size mmsize * -8
%endif
%endif
%if %2 == 8 ; chroma
cglobal vp8_%1_loop_filter8uv_mbedge, 6, 6, 15, stack_size, dst1, dst8, stride, flimE, flimI, hevthr
%else ; luma
cglobal vp8_%1_loop_filter16y_mbedge, 5, 5, 15, stack_size, dst1, stride, flimE, flimI, hevthr
%endif
%if cpuflag(ssse3)
pxor m7, m7
%endif
%ifndef m8
; splat function arguments
SPLATB_REG m0, flimEq, m7 ; E
SPLATB_REG m1, flimIq, m7 ; I
SPLATB_REG m2, hevthrq, m7 ; hev_thresh
%define m_flimE [rsp]
%define m_flimI [rsp+mmsize]
%define m_hevthr [rsp+mmsize*2]
%define m_maskres [rsp+mmsize*3]
%define m_limres [rsp+mmsize*4]
%define m_p0backup [rsp+mmsize*3]
%define m_q0backup [rsp+mmsize*4]
%define m_p2backup [rsp+mmsize*5]
%define m_q2backup [rsp+mmsize*6]
%if mmsize == 16
%define m_limsign [rsp]
%else
%define m_limsign [rsp+mmsize*7]
%endif
mova m_flimE, m0
mova m_flimI, m1
mova m_hevthr, m2
%else ; sse2 on x86-64
%define m_flimE m9
%define m_flimI m10
%define m_hevthr m11
%define m_maskres m12
%define m_limres m8
%define m_p0backup m12
%define m_q0backup m8
%define m_p2backup m13
%define m_q2backup m14
%define m_limsign m9
; splat function arguments
SPLATB_REG m_flimE, flimEq, m7 ; E
SPLATB_REG m_flimI, flimIq, m7 ; I
SPLATB_REG m_hevthr, hevthrq, m7 ; hev_thresh
%endif
%if %2 == 8 ; chroma
DEFINE_ARGS dst1, dst8, mstride, stride, dst2
%else
DEFINE_ARGS dst1, mstride, stride, dst2, dst8
%endif
mov strideq, mstrideq
neg mstrideq
%ifidn %1, h
lea dst1q, [dst1q+strideq*4-4]
%if %2 == 8 ; chroma
lea dst8q, [dst8q+strideq*4-4]
%endif
%endif
; read
lea dst2q, [dst1q+ strideq ]
%ifidn %1, v
%if %2 == 8 && mmsize == 16
%define movrow movh
%else
%define movrow mova
%endif
movrow m0, [dst1q+mstrideq*4] ; p3
movrow m1, [dst2q+mstrideq*4] ; p2
movrow m2, [dst1q+mstrideq*2] ; p1
movrow m5, [dst2q] ; q1
movrow m6, [dst2q+ strideq ] ; q2
movrow m7, [dst2q+ strideq*2] ; q3
%if mmsize == 16 && %2 == 8
movhps m0, [dst8q+mstrideq*4]
movhps m2, [dst8q+mstrideq*2]
add dst8q, strideq
movhps m1, [dst8q+mstrideq*4]
movhps m5, [dst8q]
movhps m6, [dst8q+ strideq ]
movhps m7, [dst8q+ strideq*2]
add dst8q, mstrideq
%endif
%else ; h
%if %2 == 16
lea dst8q, [dst1q+ strideq*8 ]
%endif
; read 16 rows of 8px each, interleave
movh m0, [dst1q+mstrideq*4]
movh m1, [dst8q+mstrideq*4]
movh m2, [dst1q+mstrideq*2]
movh m5, [dst8q+mstrideq*2]
movh m3, [dst1q+mstrideq ]
movh m6, [dst8q+mstrideq ]
movh m4, [dst1q]
movh m7, [dst8q]
punpcklbw m0, m1 ; A/I
punpcklbw m2, m5 ; C/K
punpcklbw m3, m6 ; D/L
punpcklbw m4, m7 ; E/M
add dst8q, strideq
movh m1, [dst2q+mstrideq*4]
movh m6, [dst8q+mstrideq*4]
movh m5, [dst2q]
movh m7, [dst8q]
punpcklbw m1, m6 ; B/J
punpcklbw m5, m7 ; F/N
movh m6, [dst2q+ strideq ]
movh m7, [dst8q+ strideq ]
punpcklbw m6, m7 ; G/O
; 8x16 transpose
TRANSPOSE4x4B 0, 1, 2, 3, 7
%ifdef m8
SWAP 1, 8
%else
mova m_q0backup, m1
%endif
movh m7, [dst2q+ strideq*2]
movh m1, [dst8q+ strideq*2]
punpcklbw m7, m1 ; H/P
TRANSPOSE4x4B 4, 5, 6, 7, 1
SBUTTERFLY dq, 0, 4, 1 ; p3/p2
SBUTTERFLY dq, 2, 6, 1 ; q0/q1
SBUTTERFLY dq, 3, 7, 1 ; q2/q3
%ifdef m8
SWAP 1, 8
SWAP 2, 8
%else
mova m1, m_q0backup
mova m_q0backup, m2 ; store q0
%endif
SBUTTERFLY dq, 1, 5, 2 ; p1/p0
%ifdef m12
SWAP 5, 12
%else
mova m_p0backup, m5 ; store p0
%endif
SWAP 1, 4
SWAP 2, 4
SWAP 6, 3
SWAP 5, 3
%endif
; normal_limit for p3-p2, p2-p1, q3-q2 and q2-q1
mova m4, m1
SWAP 4, 1
psubusb m4, m0 ; p2-p3
psubusb m0, m1 ; p3-p2
por m0, m4 ; abs(p3-p2)
mova m4, m2
SWAP 4, 2
psubusb m4, m1 ; p1-p2
mova m_p2backup, m1
psubusb m1, m2 ; p2-p1
por m1, m4 ; abs(p2-p1)
mova m4, m6
SWAP 4, 6
psubusb m4, m7 ; q2-q3
psubusb m7, m6 ; q3-q2
por m7, m4 ; abs(q3-q2)
mova m4, m5
SWAP 4, 5
psubusb m4, m6 ; q1-q2
mova m_q2backup, m6
psubusb m6, m5 ; q2-q1
por m6, m4 ; abs(q2-q1)
pmaxub m0, m1
pmaxub m6, m7
pmaxub m0, m6
; normal_limit and high_edge_variance for p1-p0, q1-q0
SWAP 7, 3 ; now m7 is zero
%ifidn %1, v
movrow m3, [dst1q+mstrideq ] ; p0
%if mmsize == 16 && %2 == 8
movhps m3, [dst8q+mstrideq ]
%endif
%elifdef m12
SWAP 3, 12
%else
mova m3, m_p0backup
%endif
mova m1, m2
SWAP 1, 2
mova m6, m3
SWAP 3, 6
psubusb m1, m3 ; p1-p0
psubusb m6, m2 ; p0-p1
por m1, m6 ; abs(p1-p0)
pmaxub m0, m1 ; max_I
SWAP 1, 4 ; max_hev_thresh
SWAP 6, 4 ; now m6 is I
%ifidn %1, v
movrow m4, [dst1q] ; q0
%if mmsize == 16 && %2 == 8
movhps m4, [dst8q]
%endif
%elifdef m8
SWAP 4, 8
%else
mova m4, m_q0backup
%endif
mova m1, m4
SWAP 1, 4
mova m7, m5
SWAP 7, 5
psubusb m1, m5 ; q0-q1
psubusb m7, m4 ; q1-q0
por m1, m7 ; abs(q1-q0)
pxor m7, m7
pmaxub m0, m1
pmaxub m6, m1
psubusb m0, m_flimI
psubusb m6, m_hevthr
pcmpeqb m0, m7 ; max(abs(..)) <= I
pcmpeqb m6, m7 ; !(max(abs..) > thresh)
%ifdef m12
SWAP 6, 12
%else
mova m_maskres, m6 ; !(abs(p1-p0) > hev_t || abs(q1-q0) > hev_t)
%endif
; simple_limit
mova m1, m3
SWAP 1, 3
mova m6, m4 ; keep copies of p0/q0 around for later use
SWAP 6, 4
psubusb m1, m4 ; p0-q0
psubusb m6, m3 ; q0-p0
por m1, m6 ; abs(q0-p0)
paddusb m1, m1 ; m1=2*abs(q0-p0)
mova m7, m2
SWAP 7, 2
mova m6, m5
SWAP 6, 5
psubusb m7, m5 ; p1-q1
psubusb m6, m2 ; q1-p1
por m7, m6 ; abs(q1-p1)
pxor m6, m6
pand m7, [pb_FE]
psrlq m7, 1 ; abs(q1-p1)/2
paddusb m7, m1 ; abs(q0-p0)*2+abs(q1-p1)/2
psubusb m7, m_flimE
pcmpeqb m7, m6 ; abs(q0-p0)*2+abs(q1-p1)/2 <= E
pand m0, m7 ; normal_limit result
; filter_common; at this point, m2-m5=p1-q1 and m0 is filter_mask
%ifdef m8 ; x86-64 && sse2
mova m8, [pb_80]
%define m_pb_80 m8
%else ; x86-32 or mmx/mmxext
%define m_pb_80 [pb_80]
%endif
mova m1, m4
mova m7, m3
pxor m1, m_pb_80
pxor m7, m_pb_80
psubsb m1, m7 ; (signed) q0-p0
mova m6, m2
mova m7, m5
pxor m6, m_pb_80
pxor m7, m_pb_80
psubsb m6, m7 ; (signed) p1-q1
mova m7, m_maskres
paddsb m6, m1
paddsb m6, m1
paddsb m6, m1
pand m6, m0
%ifdef m8
mova m_limres, m6 ; 3*(qp-p0)+(p1-q1) masked for filter_mbedge
pand m_limres, m7
%else
mova m0, m6
pand m0, m7
mova m_limres, m0
%endif
pandn m7, m6 ; 3*(q0-p0)+(p1-q1) masked for filter_common
mova m1, [pb_F8]
mova m6, m7
paddsb m7, [pb_3]
paddsb m6, [pb_4]
pand m7, m1
pand m6, m1
pxor m1, m1
pxor m0, m0
pcmpgtb m1, m7
psubb m0, m7
psrlq m7, 3 ; +f2
psrlq m0, 3 ; -f2
pand m0, m1
pandn m1, m7
psubusb m3, m0
paddusb m3, m1 ; p0+f2
pxor m1, m1
pxor m0, m0
pcmpgtb m0, m6
psubb m1, m6
psrlq m6, 3 ; +f1
psrlq m1, 3 ; -f1
pand m1, m0
pandn m0, m6
psubusb m4, m0
paddusb m4, m1 ; q0-f1
; filter_mbedge (m2-m5 = p1-q1; lim_res carries w)
%if cpuflag(ssse3)
mova m7, [pb_1]
%else
mova m7, [pw_63]
%endif
%ifdef m8
SWAP 1, 8
%else
mova m1, m_limres
%endif
pxor m0, m0
mova m6, m1
pcmpgtb m0, m1 ; which are negative
%if cpuflag(ssse3)
punpcklbw m6, m7 ; interleave with "1" for rounding
punpckhbw m1, m7
%else
punpcklbw m6, m0 ; signed byte->word
punpckhbw m1, m0
%endif
mova m_limsign, m0
%if cpuflag(ssse3)
mova m7, [pb_27_63]
%ifndef m8
mova m_limres, m1
%endif
%ifdef m10
SWAP 0, 10 ; don't lose lim_sign copy
%endif
mova m0, m7
pmaddubsw m7, m6
SWAP 6, 7
pmaddubsw m0, m1
SWAP 1, 0
%ifdef m10
SWAP 0, 10
%else
mova m0, m_limsign
%endif
%else
mova m_maskres, m6 ; backup for later in filter
mova m_limres, m1
pmullw m6, [pw_27]
pmullw m1, [pw_27]
paddw m6, m7
paddw m1, m7
%endif
psraw m6, 7
psraw m1, 7
packsswb m6, m1 ; a0
pxor m1, m1
psubb m1, m6
pand m1, m0 ; -a0
pandn m0, m6 ; +a0
%if cpuflag(ssse3)
mova m6, [pb_18_63] ; pipelining
%endif
psubusb m3, m1
paddusb m4, m1
paddusb m3, m0 ; p0+a0
psubusb m4, m0 ; q0-a0
%if cpuflag(ssse3)
SWAP 6, 7
%ifdef m10
SWAP 1, 10
%else
mova m1, m_limres
%endif
mova m0, m7
pmaddubsw m7, m6
SWAP 6, 7
pmaddubsw m0, m1
SWAP 1, 0
%ifdef m10
SWAP 0, 10
%endif
mova m0, m_limsign
%else
mova m6, m_maskres
mova m1, m_limres
pmullw m6, [pw_18]
pmullw m1, [pw_18]
paddw m6, m7
paddw m1, m7
%endif
mova m0, m_limsign
psraw m6, 7
psraw m1, 7
packsswb m6, m1 ; a1
pxor m1, m1
psubb m1, m6
pand m1, m0 ; -a1
pandn m0, m6 ; +a1
%if cpuflag(ssse3)
mova m6, [pb_9_63]
%endif
psubusb m2, m1
paddusb m5, m1
paddusb m2, m0 ; p1+a1
psubusb m5, m0 ; q1-a1
%if cpuflag(ssse3)
SWAP 6, 7
%ifdef m10
SWAP 1, 10
%else
mova m1, m_limres
%endif
mova m0, m7
pmaddubsw m7, m6
SWAP 6, 7
pmaddubsw m0, m1
SWAP 1, 0
%else
%ifdef m8
SWAP 6, 12
SWAP 1, 8
%else
mova m6, m_maskres
mova m1, m_limres
%endif
pmullw m6, [pw_9]
pmullw m1, [pw_9]
paddw m6, m7
paddw m1, m7
%endif
%ifdef m9
SWAP 7, 9
%else
mova m7, m_limsign
%endif
psraw m6, 7
psraw m1, 7
packsswb m6, m1 ; a1
pxor m0, m0
psubb m0, m6
pand m0, m7 ; -a1
pandn m7, m6 ; +a1
%ifdef m8
SWAP 1, 13
SWAP 6, 14
%else
mova m1, m_p2backup
mova m6, m_q2backup
%endif
psubusb m1, m0
paddusb m6, m0
paddusb m1, m7 ; p1+a1
psubusb m6, m7 ; q1-a1
; store
%ifidn %1, v
movrow [dst2q+mstrideq*4], m1
movrow [dst1q+mstrideq*2], m2
movrow [dst1q+mstrideq ], m3
movrow [dst1q], m4
movrow [dst2q], m5
movrow [dst2q+ strideq ], m6
%if mmsize == 16 && %2 == 8
add dst8q, mstrideq
movhps [dst8q+mstrideq*2], m1
movhps [dst8q+mstrideq ], m2
movhps [dst8q], m3
add dst8q, strideq
movhps [dst8q], m4
movhps [dst8q+ strideq ], m5
movhps [dst8q+ strideq*2], m6
%endif
%else ; h
inc dst1q
inc dst2q
; 4x8/16 transpose
TRANSPOSE4x4B 1, 2, 3, 4, 0
SBUTTERFLY bw, 5, 6, 0
lea dst8q, [dst8q+mstrideq+1]
WRITE_4x4D 1, 2, 3, 4, dst1q, dst2q, dst8q, mstrideq, strideq, %2
lea dst1q, [dst2q+mstrideq+4]
lea dst8q, [dst8q+mstrideq+4]
%if cpuflag(sse4)
add dst2q, 4
%endif
WRITE_8W m5, dst2q, dst1q, mstrideq, strideq
%if cpuflag(sse4)
lea dst2q, [dst8q+ strideq ]
%endif
WRITE_8W m6, dst2q, dst8q, mstrideq, strideq
%endif
RET
%endmacro
INIT_XMM sse2
MBEDGE_LOOPFILTER v, 16
MBEDGE_LOOPFILTER h, 16
MBEDGE_LOOPFILTER v, 8
MBEDGE_LOOPFILTER h, 8
INIT_XMM ssse3
MBEDGE_LOOPFILTER v, 16
MBEDGE_LOOPFILTER h, 16
MBEDGE_LOOPFILTER v, 8
MBEDGE_LOOPFILTER h, 8
INIT_XMM sse4
MBEDGE_LOOPFILTER h, 16
MBEDGE_LOOPFILTER h, 8
|
